take over previous run from another user

[Janpieter_Sollie]

I have a previous run from a user running under my username 'test' I'd like to take over as 'foldingathome' user (where its home dir is at /opt/foldingathome/).
The old user used the GPU, the new user uses 2 GPUs and 1 CPU.
What I tried so far:

Code: Select all

mv cores/cores.foldingathome.org/v7/lin/64bit/Core_22.fah /opt/foldingathome/cores/cores.foldingathome.org/v7/lin/64bit/
mv cores/cores.foldingathome.org/v7/lin/64bit/avx /opt/foldingathome/cores/cores.foldingathome.org/v7/lin/64bit/
mv work/01 /opt/foldingathome/work/

...but the GPU job does not move
any suggestions?

[Joe_H]

The WUs and the client are not set up to allow easy movement from on client to another. If the client did not download the WU, it does not recognize the WU and will not process it.

[Janpieter_Sollie]

OK, then what do you suggest?
I'd really like to keep f@h running under "foldingathome" in order to have better resource management, and in meantime, it also picked up WUs itself, so how do I decline the WUs attached to test?

[NoMoreQuarantine]

You can click the "Finish" button in the Advanced Control window to tell it to not request another WU after the current WU is complete. I am not aware of any way to decline a WU that has already been accepted unfortunately.

[PantherX]

...I'd really like to keep f@h running under "foldingathome" in order to have better resource management...

Welcome to the F@H Forum Janpieter_Sollie,

Could you please elaborate on that a bit more? What's the use case or ideal state that you would F@H to run in on your system?

[Janpieter_Sollie]

well, I have a kind of PC I use for NAS storage, but it was initially built for machine leaning and opencl programming. But currently, I'm not doing that (lockdown etc).
So, as the device is a good device to run heavy calculations (2 vega 64 GPUs and 1 threadripper) it could be useful to commit it to research.
But, in order to save my electricity invoice, and to protect CPU/GPUs from overheating, I'd like to configure the linux kernel to force a halt of all CPU tasks ran by folding@home when the CPU / GPU becomes too hot, or when the energy consumption monitor says the amount of energy I want to spend daily is used.

[PantherX]

If your CPU/GPU are overheating, it would be a cooling system issue. Ensure that there's plenty of air flow and that you're using appropriate cooling. If you need guidance, you can ask here or search the internet.

Regarding the power usage, my suggestion would be to experiment with the CPU/GPU slots to find a setup which you can run for most of the time within your daily cost. Say your daily cost is $10 so you test with only 1 GPU folding. If you have used up your $10 in less than 24 hours, then GPU folding might not be suitable for you. Instead, you try folding on 8 CPUs and see if that uses $10 within 24 hours. If let's say you used $7 in 24 hours while folding, then bump up the CPUs to 12 and see if that uses $10. If it does, you have found a setup which meets your power allowance and most likely your cooling target too since you're not using all the CPU/GPU simultaneously.

Reason to have a preference to find a setup that runs 24/7 is that it would be less hassle for you (no need to monitor) and you also get more points since the WUs would be returned faster as opposed to folding and have having to wait until the next day. Moreover, since each WU is unique, there's no duplication of work happening. Do note that if you want to get bonus points, you need to have a passkey.

[Janpieter_Sollie]

what are bonus points? I have a key, but no idea what they are.

Actually, I wanted to make it more efficient than that: GPUs are not always used, and during night, the electricity is around 50% less expensive. I'd like to run any WU I can get during night, and afterwards run the calculations which could still be done at a higher cost. I know, it's a little bit crazy, but I thought it is the cheapest way to do this.
Yes, this PC has a cooling problem: many medium towers are simply not made to host 800W TDP (GPU+CPU+HDDs) for a long time. I don't have room for a big tower, so here's your cause.
No, I won't monitor it: we got hardware sensors to do that, and shell scripting to calculate the appropriate actions...
I really appreciate the fact that you want to point me to an easy solution, but just grant me the challenge . One day, it may even become useful to others

[PantherX]

what are bonus points? I have a key, but no idea what they are...

Here you go: https://foldingathome.org/support/faq/points/passkey/

...Actually, I wanted to make it more efficient than that: GPUs are not always used, and during night, the electricity is around 50% less expensive. I'd like to run any WU I can get during night, and afterwards run the calculations which could still be done at a higher cost. I know, it's a little bit crazy, but I thought it is the cheapest way to do this...

It's perfectly fine to customize F@H to suit your needs! A quick way would be that you configure the CPU and GPU slots to fold only when idle (the detection of idle will vary from OS to OS). Thus, at night-time, when the system is idle, it will fold and during the day, when you're working, it won't fold. It will fold if the system is idle during the day but it won't be for very long if you're using it regularly. See if that helps you out. Also, if you need help figuring out what idle state is, provide us with your OS details and we can help you with that.

...I really appreciate the fact that you want to point me to an easy solution, but just grant me the challenge . One day, it may even become useful to others

We like to keep information public so that it help others. Sometime, others can help you too so there's a lot of benefit of having public Forums

[Joe_H]

The client can also be communicated with over a telnet port, local IP address 127.0.0.1 on port 36330. People have used that to set up scripts to have the client do various things such as pause or start at certain times. That might be something you would be interested in using, for instance have the client start folding on a GPU at certain times, and pause or Finish at other times.